Exploring Deep Neural Networks to Retrieve Rain and Snow in High Latitudes Using Multisensor and Reanalysis Data
AffiliationUniv Arizona, Dept Hydrol & Atmospher Sci
MetadataShow full item record
PublisherAMER GEOPHYSICAL UNION
CitationTang, G., Long, D., Behrangi, A., Wang, C., & Hong, Y. ( 2018). Exploring deep neural networks to retrieve rain and snow in high latitudes using multisensor and reanalysis data. Water Resources Research, 54, 8253– 8278. https://doi.org/10.1029/2018WR023830
JournalWATER RESOURCES RESEARCH
Rights©2018. American Geophysical Union. All Rights Reserved.
Collection InformationThis item from the UA Faculty Publications collection is made available by the University of Arizona with support from the University of Arizona Libraries. If you have questions, please contact us at email@example.com.
AbstractSatellite remote sensing is able to provide information on global rain and snow, but challenges remain in accurate estimation of precipitation rates, particularly in snow retrieval. In this work, the deep neural network (DNN) is applied to estimate rain and snow rates in high latitudes. The reference data for DNN training are provided by two spaceborne radars onboard the Global Precipitation Measurement (GPM) Core Observatory and CloudSat. Passive microwave data from the GPM Microwave Imager (GMI), infrared data from MODerate resolution Imaging Spectroradiometer and environmental data from European Centre for Medium-Range Weather Forecasts are trained to the spaceborne radar-based reference precipitation. The DNN estimates are compared to data from the Goddard Profiling Algorithm (GPROF), which is used to retrieve passive microwave precipitation for the GPM mission. First, the DNN-based retrieval method performs well in both training and testing periods. Second, the DNN can reveal the advantages and disadvantages of different channels of GMI and MODerate resolution Imaging Spectroradiometer. Additionally, infrared and environmental data can improve precipitation estimation of the DNN, particularly for snowfall. Finally, based on the optimized DNN, rain and snow are estimated in 2017 from orbital GMI brightness temperatures and compared to ERA-Interim and Modern-Era Retrospective analysis for Research and Applications Version 2 reanalysis data. Evaluation results show that (1) the DNN can largely mitigate the underestimation of precipitation rates in high latitudes by GPROF; (2) the DNN-based snowfall estimates largely outperform those of GPROF; and (3) the spatial distributions of DNN-based precipitation are closer to reanalysis data. The method and assessment presented in this study could potentially contribute to the substantial improvement of satellite precipitation products in high latitudes. Plain Language Summary Snow has a significant influence on the hydrological cycle and water resource availability. Compared to ground gauges and radars with limited coverage, satellite remote sensing can provide global rain and snow observations from space. However, traditional satellite precipitation retrieval methods are prone to errors in snow estimation at high latitudes. In this study, we developed a new rain and snow estimation method at high latitudes using deep neural networks. The reference data sets were from two spaceborne radars that provide the most direct precipitation observations from space. Passive microwave, infrared, and environmental data were trained to the reference data sets for rain and snow estimation. Results show that the neural network-based method can largely reduce the underestimation of rain and snow rates in high latitudes in many prior algorithms. Statistical indices for snow estimation are notably improved. Furthermore, combining data from passive microwave, infrared, and environmental data sets contribute to better precipitation estimation than a single source. We suggest that deep neural networks could potentially contribute to the improvement of satellite precipitation at high latitudes, which is valuable for expanding the spatial coverage of current satellite products.
Note6 month embargo; published online: 11 October 2018
VersionFinal published version
SponsorsNational Natural Science Foundation of China [91547210, 71461010701, 91437214]; National Key Research and Development Program of China [2016YFE0102400]; NASA Energy and Water Cycle Study [NNH13ZDA001N-NEWS]; NASA weather program [NNH13ZDA001NWeather]; China Scholarship Council (CSC)